Changelog for package tf2

0.7.7 (2023-10-13)

• fix extra comma that gives annoying build warnings with -Wall and -Wpedantic with g++-9 and assuming most other compilers (#550)
• Add parent frame to warning logs (#533)
• Contributors: Jack Zender, Stephan

0.7.6 (2022-10-11)

• Fix dead loop in message filter (#532)
• Restore time difference order so future extrapolation exceptions don't show non-sensical negative seconds into the future (#522)
• Contributors: Feng Zhaolin, Lucas Walter

0.7.5 (2020-09-01)

• restore buffer sizes
• Contributors: Tully Foote

0.7.4 (2020-09-01)

• Fix potential buffer overrun of snprintf (#479)
• Contributors: Atsushi Watanabe

0.7.3 (2020-08-25)

• Use snprintf instead of stringstream to increase performance of lookupTransform() in error cases.
• Do not waste time constructing error string if nobody is interested in it in canTransform(). (#469)
• Output time difference of extrapolation exceptions (#477)
• Cherry-picking various commits from Melodic (#471)
  • Revert "rework Eigen functions namespace hack" (#436)

  * Fixed warnings in message_filter.h (#434) the variables are not used in function body and caused -Wunused-parameter to trigger with -Wall

  • Fix ambiguous call for tf2::convert on MSVC (#444)
  • rework ambiguous call on MSVC.
• Contributors: Lucas Walter, Martin Pecka, Robert Haschke

0.7.2 (2020-06-08)

0.7.1 (2020-05-13)

• Fix to improper ring_45 test, where 'anchor' frame for both inverse and normal transform was frame 'b' instead of frame 'a', thus creating a problem
• Don't insert a TF frame is one of the same timestamp already exists, instead just overwrite it.
• [Noetic] Add tf2::Stamped<T>::operator=() to fix warnings downstream (#453)
  • Add tf2::Stamped<T>::operator=()
• [noetic] cherry-pick Windows fixes from melodic-devel (#450)
  • [Windows][melodic-devel] Fix install locations (#442)
  • fixed install locations of tf2
  • [windows][melodic] more portable fixes. (#443)
  • more portable fixes.
• Contributors: Patrick Beeson, Robert Haschke, Sean Yen, Shane Loretz

0.7.0 (2020-03-09)

• Bump CMake version to avoid CMP0048 warning (#445) Signed-off-by: Shane Loretz <<sloretz@osrfoundation.org>>

• Fix compile error missing ros/ros.h (#400) * ros/ros.h -> ros/time.h tf2 package depends on rostime * tf2_bullet doesn't need ros.h Signed-off-by: Shane Loretz <<sloretz@osrfoundation.org>> * tf2_eigen doesn't need ros/ros.h Signed-off-by: Shane Loretz <<sloretz@osrfoundation.org>>

• Merge pull request #367 from kejxu/add_tf2_namespace_to_avoid_name_collision rework Eigen functions namespace hack
• separate transform function declarations into transform_functions.h
• use ROS_DEPRECATED macro for portability (#362)
  • use ROS_DEPRECATED for better portability
  • change ROS_DEPRECATED position (#5)
• Remove [signals]{.title-ref} from find_package(Boost COMPONENTS ...). tf2 is using signals2, which is not the same library. Additionally, signals2 has always been header only, and header only libraries must not be listed in find_package. Boost 1.69 removed the old signals library entirely, so the otherwise useless [COMPONENTS signals]{.title-ref} actually breaks the build.
• Remove legacy inclusion in CMakeLists of tf2.
• Contributors: James Xu, Maarten de Vries, Marco Tranzatto, Shane Loretz, Tully Foote

0.6.5 (2018-11-16)

0.6.4 (2018-11-06)

• Resolved pedantic warnings
• fix issue #315
• fixed nan interpoaltion issue
• Contributors: Keller Fabian Rudolf (CC-AD/EYC3), Kuang Fangjun, Martin Ganeff

0.6.3 (2018-07-09)

• preserve constness of const argument to avoid warnings (#307)
• Change comment style for unused doxygen (#297)
• Contributors: Jacob Perron, Tully Foote

0.6.2 (2018-05-02)

0.6.1 (2018-03-21)

• Replaced deprecated console_bridge macro calls (tests)
• Contributors: Johannes Meyer, Tully Foote

0.6.0 (2018-03-21)

• Replaced deprecated log macro calls
• Contributors: Tim Rakowski, Tully Foote

0.5.17 (2018-01-01)

• Merge pull request #278 from ros/chain_as_vec_test2 Clean up results of _chainAsVector
• Simple test to check BufferCore::_chainAsVector. Unit tests for walk and chain passing now.
• Merge pull request #267 from at-wat/speedup-timecache-for-large-buffer Speed-up TimeCache search for large cache time.
• Merge pull request #265 from vsherrod/interpolation_fix Corrected time output on interpolation function.
• Add time_interval option to tf2 speed-test.
• Merge pull request #269 from ros/frames_as_yaml allFrameAsYaml consistently outputting a dict
• resolve https://github.com/ros/geometry/pull/153 at the source instead of needing the workaround.
• Speed-up TimeCache search for large cache time.
• Modified tests for correct time in interpolation to existing tests.
• Corrected time output on interpolation function. Added unit test to check for this.
• Contributors: Atsushi Watanabe, Miguel Prada, Tully Foote, Vallan Sherrod

0.5.16 (2017-07-14)

• remove explicit templating to standardize on overloading. But provide backwards compatibility with deprecation.
• Merge pull request #144 from clearpathrobotics/dead_lock_fix Solve a bug that causes a deadlock in MessageFilter
• Resolve 2 places where the error_msg would not be propogated. Fixes #198
• Remove generate_rand_vectors() from a number of tests. (#227)
• fixing include directory order to support overlays (#231)
• replaced dependencies on tf2_msgs_gencpp by exported dependencies
• Document the lifetime of the returned reference for getFrameId getTimestamp
• relax normalization tolerance. #196 was too strict for some use cases. (#220)
• Solve a bug that causes a deadlock in MessageFilter
• Contributors: Adel Fakih, Chris Lalancette, Christopher Wecht, Tully Foote, dhood

0.5.15 (2017-01-24)

0.5.14 (2017-01-16)

• fixes #194 check for quaternion normalization before inserting into storage (#196)
  • check for quaternion normalization before inserting into storage
  • Add test to check for transform failure on invalid quaternion input
• updating getAngleShortestPath() (#187)
• Move internal cache functions into a namespace Fixes https://github.com/ros/geometry2/issues/175
• Link properly to convert.h
• Landing page for tf2 describing the conversion interface
• Fix comment on BufferCore::MAX_GRAPH_DEPTH.
• Contributors: Jackie Kay, Phil Osteen, Tully Foote, alex, gavanderhoorn

0.5.13 (2016-03-04)

0.5.12 (2015-08-05)

• add utilities to get yaw, pitch, roll and identity transform
• provide more conversions between types The previous conversion always assumed that it was converting a non-message type to a non-message type. Now, one, both or none can be a message or a non-message.
• Contributors: Vincent Rabaud

0.5.11 (2015-04-22)

0.5.10 (2015-04-21)

• move lct_cache into function local memoryfor #92
• Clean up range checking. Re: #92
• Fixed chainToVector
• release lock before possibly invoking user callbacks. Fixes #91
• Contributors: Jackie Kay, Tully Foote

0.5.9 (2015-03-25)

• fixing edge case where two no frame id lookups matched in getLatestCommonTime
• Contributors: Tully Foote

0.5.8 (2015-03-17)

• change from default argument to overload to avoid linking issue #84
• remove useless Makefile files
• Remove unused assignments in max/min functions
• change _allFramesAsDot() -> _allFramesAsDot(double current_time)
• Contributors: Jon Binney, Kei Okada, Tully Foote, Vincent Rabaud

0.5.7 (2014-12-23)

0.5.6 (2014-09-18)

0.5.5 (2014-06-23)

• convert to use console bridge from upstream debian package https://github.com/ros/rosdistro/issues/4633
• Fix format string
• Contributors: Austin, Tully Foote

0.5.4 (2014-05-07)

• switch to boost signals2 following ros/ros_comm#267, blocking ros/geometry#23
• Contributors: Tully Foote

0.5.3 (2014-02-21)

0.5.2 (2014-02-20)

0.5.1 (2014-02-14)

0.5.0 (2014-02-14)

0.4.10 (2013-12-26)

• updated error message. fixes #38
• tf2: add missing console bridge include directories (fix #48)
• Fix const correctness of tf2::Vector3 rotate() method The method does not modify the class thus should be const. This has already been fixed in Bullet itself.
• Contributors: Dirk Thomas, Timo Rohling, Tully Foote

0.4.9 (2013-11-06)

0.4.8 (2013-11-06)

• moving python documentation to tf2_ros from tf2 to follow the code
• removing legacy rospy dependency. implementation removed in 0.4.0 fixes #27

0.4.7 (2013-08-28)

• switching to use allFramesAsStringNoLock inside of getLatestCommonTime and walkToParent and locking in public API _getLatestCommonTime instead re #23
• Fixes a crash in tf's view_frames related to dot code generation in allFramesAsDot

0.4.6 (2013-08-28)

• cleaner fix for #19
• fix pointer initialization. Fixes #19
• fixes #18 for hydro
• package.xml: corrected typo in description

0.4.5 (2013-07-11)

• adding _chainAsVector method for https://github.com/ros/geometry/issues/18
• adding _allFramesAsDot for backwards compatability https://github.com/ros/geometry/issues/18

0.4.4 (2013-07-09)

• making repo use CATKIN_ENABLE_TESTING correctly and switching rostest to be a test_depend with that change.
• tf2: Fixes a warning on OS X, but generally safer Replaces the use of pointers with shared_ptrs, this allows the polymorphism and makes it so that the compiler doesn't yell at us about calling delete on a class with a public non-virtual destructor.
• tf2: Fixes compiler warnings on OS X This exploited a gcc specific extension and is not C++ standard compliant. There used to be a "fix" for OS X which no longer applies. I think it is ok to use this as an int instead of a double, but another way to fix it would be to use a define.
• tf2: Fixes linkedit errors on OS X

0.4.3 (2013-07-05)

0.4.2 (2013-07-05)

• adding getCacheLength() to parallel old tf API
• removing legacy static const variable MAX_EXTRAPOLATION_DISTANCE copied from tf unnecessesarily

0.4.1 (2013-07-05)

• adding old style callback notifications to BufferCore to enable backwards compatability of message filters
• exposing dedicated thread logic in BufferCore and checking in Buffer
• more methods to expose, and check for empty cache before getting latest timestamp
• adding methods to enable backwards compatability for passing through to tf::Transformer

0.4.0 (2013-06-27)

• splitting rospy dependency into tf2_py so tf2 is pure c++ library.
• switching to console_bridge from rosconsole
• moving convert methods back into tf2 because it does not have any ros dependencies beyond ros::Time which is already a dependency of tf2
• Cleaning up unnecessary dependency on roscpp
• Cleaning up packaging of tf2 including: removing unused nodehandle fixing overmatch on search and replace cleaning up a few dependencies and linking removing old backup of package.xml making diff minimally different from tf version of library
• suppressing bullet LinearMath copy inside of tf2, so it will not collide, and should not be used externally.
• Restoring test packages and bullet packages. reverting 3570e8c42f9b394ecbfd9db076b920b41300ad55 to get back more of the packages previously implemented reverting 04cf29d1b58c660fdc999ab83563a5d4b76ab331 to fix #7
• fixing includes in unit tests
• Make PythonLibs find_package python2 specific On systems with python 3 installed and default, find_package(PythonLibs) will find the python 3 paths and libraries. However, the c++ include structure seems to be different in python 3 and tf2 uses includes that are no longer present or deprecated. Until the includes are made to be python 3 compliant, we should specify that the version of python found must be python 2.

0.3.6 (2013-03-03)

0.3.5 (2013-02-15 14:46)

• 0.3.4 -> 0.3.5

0.3.4 (2013-02-15 13:14)

• 0.3.3 -> 0.3.4
• moving LinearMath includes to include/tf2

0.3.3 (2013-02-15 11:30)

• 0.3.2 -> 0.3.3
• fixing include installation of tf2

0.3.2 (2013-02-15 00:42)

• 0.3.1 -> 0.3.2
• fixed missing include export & tf2_ros dependecy

0.3.1 (2013-02-14)

• 0.3.0 -> 0.3.1
• fixing PYTHON installation directory

0.3.0 (2013-02-13)

• switching to version 0.3.0
• adding setup.py to tf2 package
• fixed tf2 exposing python functionality
• removed line that was killing tf2_ros.so
• fixing catkin message dependencies
• removing packages with missing deps
• adding missing package.xml
• adding missing package.xml
• adding missing package.xml
• catkinizing geometry-experimental
• removing bullet headers from use in header files
• removing bullet headers from use in header files
• merging my recent changes
• setting child_frame_id overlooked in revision 6a0eec022be0 which fixed failing tests
• allFramesAsString public and internal methods seperated. Public method is locked, private method is not
• fixing another scoped lock
• fixing one scoped lock
• fixing test compilation
• merge
• Error message fix, ros-pkg5085
• Check if target equals to source before validation
• When target_frame == source_frame, just returns an identity transform.
• adding addition ros header includes for strictness
• Fixed optimized lookups with compound transforms
• Fixed problem in tf2 optimized branch. Quaternion multiplication order was incorrect
• fix compilation on 32-bit
• Josh fix: Final inverse transform composition (missed multiplying the sourcd->top vector by the target->top inverse orientation). b44877d2b054
• Josh change: fix first/last time case. 46bf33868e0d
• fix transform accumulation to parent
• fix parent lookup, now works on the real pr2's tree
• move the message filter to tf2_ros
• tf2::MessageFilter + tests. Still need to change it around to pass in a callback queue, since we're being triggered directly from the tf2 buffer
• Don't add the request if the transform is already available. Add some new tests
• working transformable callbacks with a simple (incomplete) test case
• first pass at a transformable callback api, not tested yet
• add interpolation cases
• fix getLatestCommonTime -- no longer returns the latest of any of the times
• Some more optimization -- allow findClosest to inline
• another minor speedup
• Minorly speed up canTransform by not requiring the full data lookup, and only looking up the parent
• Add explicit operator= so that we can see the time in it on a profile graph. Also some minor cleanup
• minor cleanup
• add 3 more cases to the speed test
• Remove use of btTransform at all from transform accumulation, since the conversion to/from is unnecessary, expensive, and can introduce floating point error
• Don't use btTransform as an intermediate when accumulating transforms, as constructing them takes quite a bit of time
• Completely remove lookupLists(). canTransform() now uses the same walking code as lookupTransform(). Also fixed a bug in the static transform publisher test
• Genericise the walk-to-top-parent code in lookupTransform so that it will be able to be used by canTransform as well (minus the cost of actually computing the transform)
• remove id lookup that wasn't doing anything
• Some more optimization:
  • Reduce # of TransformStorage copies made in TimeCache::getData()
  • Remove use of lookupLists from getLatestCommonTime
• lookupTransform() no longer uses lookupLists unless it's called with Time(0). Removes lots of object construction/destruction due to removal of pushing back on the lists
• Remove CompactFrameID in favor of a typedef
• these mode checks are no longer necessary
• Fix crash when testing extrapolation on the forward transforms
• Update cache unit tests to work with the changes TransformStorage. Also make sure that BT_USE_DOUBLE_PRECISION is set for tf2.
• remove exposure of time_cache.h from buffer_core.h
• Removed the mutex from TimeCache, as it's unnecessary (BufferCore needs to have its own mutex locked anyway), and this speeds things up by about 20% Also fixed a number of thread-safety problems
• Optimize test_extrapolation a bit, 25% speedup of lookupTransform
• use a hash map for looking up frame numbers, speeds up lookupTransform by ~8%
• Cache vectors used for looking up transforms. Speeds up lookupTransform by another 10%
• speed up lookupTransform by another 25%
• speed up lookupTransform by another 2x. also reduces the memory footprint of the cache significantly
• sped up lookupTransform by another 2x
• First add of a simple speed test Sped up lookupTransform 2x
• roscpp dependency explicit, instead of relying on implicit
• static transform tested and working
• tests passing and all throw catches removed too!
• validating frame_ids up front for lookup exceptions
• working with single base class vector
• tests passing for static storage
• making method private for clarity
• static cache implementation and test
• cleaning up API doc typos
• sphinx docs for Buffer
• new dox mainpage
• update tf2 manifest
• commenting out twist
• Changed cache_time to cache_time to follow C++ style guide, also initialized it to actually get things to work
• no more rand in cache tests
• Changing tf2_py.cpp to use underscores instead of camelCase
• removing all old converter functions from transform_datatypes.h
• removing last references to transform_datatypes.h in tf2
• transform conversions internalized
• removing unused datatypes
• copying bullet transform headers into tf2 and breaking bullet dependency
• merge
• removing dependency on tf
• removing include of old tf from tf2
• update doc
• merge
• kdl unittest passing
• Spaces instead of tabs in YAML grrrr
• Adding quotes for parent
• canTransform advanced ported
• Hopefully fixing YAML syntax
• new version of view_frames in new tf2_tools package
• testing new argument validation and catching bug
• Python support for debugging
• merge
• adding validation of frame_ids in queries with warnings and exceptions where appropriate
• Exposing ability to get frames as a string
• A compiling version of YAML debugging interface for BufferCore
• placeholder for tf debug
• fixing tf:: to tf2:: ns issues and stripping slashes on set in tf2 for backwards compatiabily
• Adding a python version of the BufferClient
• moving test to new package
• merging
• working unit test for BufferCore::lookupTransform
• removing unused method test and converting NO_PARENT test to new API
• Adding some comments
• Moving the python bindings for tf2 to the tf2 package from the tf2_py package
• buffercore tests upgraded
• porting tf_unittest while running incrmentally instead of block copy
• BufferCore::clear ported forward
• successfully changed lookupTransform advanced to new version
• switching to new implementation of lookupTransform tests still passing
• compiling lookupTransform new version
• removing tf_prefix from BufferCore. BuferCore is independent of any frame_ids. tf_prefix should be implemented at the ROS API level.
• initializing tf_prefix
• adding missing initialization
• suppressing warnings
• more tests ported
• removing tests for apis not ported forward
• setTransform tests ported
• old tests in new package passing due to backwards dependency. now for the fun, port all 1500 lines :-)
• setTransform working in new framework as well as old
• porting more methods
• more compatability
• bringing in helper functions for buffer_core from tf.h/cpp
• rethrowing to new exceptions
• converting Storage to geometry_msgs::TransformStamped
• removing deprecated useage
• cleaning up includes
• moving all implementations into cpp file
• switching test to new class from old one
• Compiling version of the buffer client
• moving listener to tf_cpp
• removing listener, it should be in another package
• most of listener
• add cantransform implementation
• removing deprecated API usage
• initial import of listener header
• move implementation into library
• 2 tests of buffer
• moving executables back into bin
• compiling again with new design
• rename tfcore to buffercore
• almost compiling version of template code
• compiling tf2_core simple test
• add test to start compiling
• copying in tf_unittest for tf_core testing template
• prototype of tf2_core implemented using old tf.
• first version of template functions
• remove timeouts
• properly naming tf2_core.h from tf_core.h
• working cache test with tf2 lib
• first unit test passing, not yet ported
• tf_core api
• tf2 v2
• aborting port
• moving across time cache tf and datatypes headers
• copying exceptions from tf
• switching to tf2 from tf_core